def build_list_and_evaluate_and_save_chart(
self, list_results: List, gt_graph: GraphDataStruct,
only_decisions: List[scoring_datastrutures.DecisionTypes],
output_folder: pathlib.Path):
# Generate name of the list
generated_name = "".join([d.name + "_"
for d in only_decisions]) + "only"
# Filter out results
results_list_filtered = [
self.filter_out_request_result(r, only_decisions)
for r in list_results
]
json_import_export.save_json(
results_list_filtered,
pathlib.Path(get_homedir() / (generated_name + ".json")))
perfs_list_filtered = self._compute_perfs_list(results_list_filtered,
gt_graph)
# Save to graph
twoDplot = two_dimensions_plot.TwoDimensionsPlot()
twoDplot.print_graph(perfs_list_filtered,
output_folder,
file_name=(generated_name + ".png"))
def get_proximity_graph_and_evaluate(self, image_folder: pathlib.Path, output_path: pathlib.Path, gt_path: pathlib.Path):
list_results, _ = self.get_proximity_graph(image_folder, output_path)
perfs_list = self.evaluate_list_results(list_results, gt_path, output_path)
# Save to file
json_import_export.save_json(perfs_list, output_path / "perfs_list.json")
self.logger.debug(f"Performance list saved.")
# Save to graph
twoDplot = two_dimensions_plot.TwoDimensionsPlot()
twoDplot.print_graph(perfs_list, output_path)
def print_data(self,
scalabilitygraph: ScalabilityData,
output_folder: pathlib.Path,
file_name: str = "scalability_graph.pdf"):
twoDplot = two_dimensions_plot.TwoDimensionsPlot()
twoDplot.print_scalability_data(scalabilitygraph, output_folder,
file_name)
# Save to file
json_import_export.save_json(scalabilitygraph.list_request_time,
output_folder / "scalability_graph.json")
self.logger.info("Results scalability_graph json saved.")
def launch(self):
# ========= INPUTS =========
# Input files folder
image_folder = get_homedir() / "datasets" / "MINI_DATASET"
# Ground truth file
gt = get_homedir() / "datasets" / "MINI_DATASET_VISJS.json"
# Output general folder
output_folder = get_homedir() / "datasets" / "OUTPUT"
output_folder.mkdir(parents=True, exist_ok=True)
# ========= GOAL =========
perfs = []
iterations_limit = 50 # Or nb of iteration if complete exploration
max_threshold = 1
min_threshold = 0
# ========= CONFIGURATION CHOSING =========
for i in range(iterations_limit):
# Computing the new threshold
curr_threshold = i * (
(max_threshold - min_threshold) / iterations_limit)
self.logger.info(
f"Current threshold computation : {curr_threshold}")
# If the instance already exist, delete it
if self.server_launcher is not None:
del self.server_launcher
# Put configuration in place
self.server_launcher = instance_handler.Instance_Handler()
self.server_launcher.dist_conf.MAX_DIST_FOR_NEW_CLUSTER = curr_threshold
# Create output folder for this configuration
tmp_output = output_folder / ''.join(
[str(curr_threshold), "_threshold"])
tmp_output.mkdir(parents=True, exist_ok=True)
# ========= CONFIGURATION LAUNCH =========
# Launch Server
self.server_launcher.launch()
time.sleep(2)
# Launch client tester
self.client_launcher = evaluator.InternalClusteringQualityEvaluator(
)
perf_overview = self.client_launcher.get_storage_graph(
image_folder, gt, tmp_output)
self.logger.warning(f"Perf overview added : {perf_overview}")
perfs.append(perf_datastruct.Perf(perf_overview, curr_threshold))
# Wait for client end
# ========= TIDY UP FOR NEXT ROUND =========
# Flush server
self.server_launcher.flush_db()
# Shutdown server
self.server_launcher.stop()
# Wait for shutdown (wait for workers to shutdown, usually longer than db)
while not self.server_launcher.check_worker():
time.sleep(1) # Enough ?
self.logger.warning("Waiting for workers to stop .. ")
# Remove all workers
self.server_launcher.flush_workers()
time.sleep(2)
# Print plot
TwoD_plot = two_dimensions_plot.TwoDimensionsPlot()
TwoD_plot.print_graph(perfs, output_folder)
def setUp(self):
self.logger = logging.getLogger()
# self.conf = .Default_configuration()
self.test_path = get_homedir(
) / "datasets" / "douglas-quaid-tests" / "TwoDPlot"
self.matrix_gen = two_dimensions_plot.TwoDimensionsPlot()